The goal of this research is to explore and to model on-line user behavior with respect to information-retrieval systems. Modeling will be based upon empirical data already routinely collected from the National Library of Medicine's on-line interactive system (MEDLINE). The research procedure will involve pattern analysis of user transactions wherein ten hypotheses will be tested regarding methods of representing, analyzing, and improving user search behavior. As a result of testing, detailed descriptions of search behaviors will be produced and variant/invariant factors in search patterns will be identified. These results will be used to evaluate effective and ineffective on-line user behavior patterns. From the research and analysis, a model of user behavior will be developed and validated for use in developing improved prompting techniques based on the concept of "adaptive", or user-sensitive, prompting and diagnostics.